The problem
Municipal tree registers are good at the things a register is for: species, location, trunk diameter, when it was last inspected. They do not record how much leaf a tree actually carries, which is the quantity that matters for shade, cooling and interception. Measuring it directly, tree by tree, across a city is not something anyone is going to fund.
What I build
Cities already fly aerial LiDAR surveys on a regular cycle for entirely unrelated reasons. Those scans contain the information: a pulse that passes through a canopy and a pulse that is intercepted by it are distinguishable, so the fraction of sky visible through the crown can be recovered from data that already exists.
From that gap fraction, leaf area follows by a physical inversion rather than a fitted model, which is the part I care about most. The relationship between how much light a canopy blocks and how much leaf it carries is classical physics with known assumptions, so the estimate can be argued about on its own terms instead of defended as the output of a regressor nobody can inspect. An earlier version of this work did use machine learning; the current method does not.
What is shown here
Nothing but the method. This is a multi-author manuscript in preparation, so there are no numbers, figures or findings on this page, and there will not be until it is published.